Skip to content

Iran vote count: “fraud pretty clear now,” sez Mebane

Update here and data here. I haven’t looked at this in detail, but Walter Mebane is the expert on this stuff so I’m inclined to believe him. Even though he uses tables instead of graphs.

Again, just to emphasize: this sort of statistical analysis doesn’t prove anything by itself, but it can be useful in giving people a sense of where to focus attention if they want to look further.


  1. Krish Swamy says:

    The point of this type of analysis beats me. It is clearly not conclusive but only is a directional read of some malfeasance in the electoral process. At this point, the analysis is purely specification.

    Also, the analysis does not improve anyone's lot in the world. Only goes to establish that the Iranian people have been potentially gypped. Seems like an open secret. So, to what end?

  2. Andrew Gelman says:

    Krish: Mebane's analysis does not solve all problems but it provides information. Information is good.

  3. anon says:

    I haven't been following this too closely, nor have I thoroughly read Mebane's stuff, but I will not believe any of it until I see the exact same methods applied to elections that have been thought to be legitimate.

  4. ChristianK says:

    Is there a place where you can get those election numbers simply as csv file or in another computer readable format to examine them for yourself?

  5. alex says:

    Andrew: I sincerely hope that you do look at this in detail and write something up. As far as statistical analyses of the Iranian data go, Walter's work seems to be the only arguably compelling game in town – his also appears to be the only analysis using ballot box level data.

    As someone with basic graduate level training in statistics, I find it difficult to evaluate his claims. Being a top notch statistician doesn't mean that his analysis is beyond question (the work its based on is quite preliminary). Is there even a compelling reason to think that Iranian polling place level data should satisfy 2nd digit benford's law (other than "well, it seems to work in some contexts, but not in others"). Walter is suitably modest about how much the statistics can say given the limited amount of data available in the vote counts, but nevertheless I'm sure a second professional opinion about what the data does say would be appreciated by a lot of people.

    Everybody seems to take it as given that fraud occurred, but nobody has offered a plausible mechanism for fraud on the scale that's claimed. Early reports specifically rejected "ballot box stuffing" as being too small potatoes to get the job done, which is the mechanism Walter seems to be proposing. It would be nice to see some discussion of the statistical claims out there from the perspective of "what is the likelihood of fraud given these analyses," rather than "we already know that there was fraud, and look somebody is telling us with data what we already know." My impression is that people are skittish about subjecting the analyses claiming fraud to the kind of rigorous criticism typically applied to most academic work (a WaPo op-ed piece comes to mind). Maybe some individuals early on had a bad experience doing so (Karroubi's 7's, anyone?) so its understandable. Nevertheless, this particular subject seems too important not to.

  6. Thomas Lotze says:

    Christian: for the election numbers, take a look at the data link here.

    alex: I agree, a discussion of plausible, testable mechanisms for generating fraud will need to be the next chapter in this story. The speculation by Beber and Scacco (posted in this same blog, I think is a good first step in this direction.

  7. FredH says:

    Mebane's analysis is done using Benford's Law. Benford's Law is occasionally used to detect financial fraud, but its use in election analysis has been thoroughly discredited. The Carter Center expert election panel of 4 professors have concluded that "there is insufficient evidence that Benfords Law applies to election results in general. Furthermore, a simple but plausible model of the election does not produce results that conform to Benfords Law." This was after the Venezuelan referendum which they officially monitored and declared fair was accused of being fraudulent by the opposition based on Benfords Law analysis. The Carter Center has monitored 76 elections in 30 countries.

    src: , see page 128 (outline) and page 132-133 (details)